Estimation of mineral content in natural grasslands by near infrared reflectance spectroscopy

Abstract

Near infrared reflectance spectroscopy (NIRS) was tested to predict nitrogen and mineral concentration (N, P, K, Ca, Mg, Na, Mn, Fe, Cu and Zn) in natural grassland samples. The samples were taken from different community types according to the topographic gradient, at different maturation stages and during a period of four consecutive years. A subset of 95 samples was selected on the basis of the spectral variation. Chemical values from the calibratíon sample set were regressed on the corresponding spectral data using a stepwise multiple regression analysís. Another subset of 75 samples was used as the validation set. Standard errors of prediction and correlation coefficients, respectively, were: 0.71 and 0.97 (N), 0.22 and 0.73 (P), 1.83 and 0.84 (K), 0.83 and 0.92 (Ca), 0.15 and 0.92 (Mg), 3.94 and 0.66 (Na), 44 and 0.84 (Mn), 19 and 0.75 (Fe), 1.01 and 0.77 (Cu), 3.9 and 0.79 (Zn).